CQS failures and structure failures

The CQS subsystem automatically recovers from most types of failures. Because it is registered with the z/OS® Automatic Restart Manager, if the CQS subsystem fails, z/OS restarts it. If one of the structures in the coupling facility fails, CQS initiates a structure rebuild (using data from the CQS log) to recover the structure.

If you lose connection with the last remaining coupling facility, IMS stops sharing the message queues and stops to work because there are no queues available for input or output. After the connection is restored, CQS reconnects to the structures and rebuilds them if necessary, and IMS resumes work.

In the case of a resource structure, a coupling facility list structure, CQS does not support structure recovery, structure checkpointing, or overflow processing. To provide optimum performance, CQS does not log changes to resource structures. To aid in recovering resources, you can define resource structures in the CFRM policy to duplex them automatically.

In the case of a nonrecoverable queue structure (RECOVERABLE=NO), CQS does not support structure recovery. When a nonrecoverable queue structure fails, CQS begins recovery processing, and cold starts the newly-allocated structure (initializes it to empty). All active CQSs that were connected to the failed structure then abend with an ABENDU0373. You should restart the CQSs, at which time, the CQSs performs resync processing with any connected clients. You can manually restart CQS, use the z/OS ARM feature (ARMRST=Y), or use your own automation. When CQS restarts after the failure of a nonrecoverable queue structure, all data in the previous (failed) structure is lost.